CHANGE DETECTION IN PHOTOGRAMMETRIC POINT ... - pf.bgu.tum… · change detection in...

7
CHANGE DETECTION IN PHOTOGRAMMETRIC POINT CLOUDS FOR MONITORING OF ALPINE, GRAVITATIONAL MASS MOVEMENTS A. Dinkel, L. Hoegner * , A. Emmert, L. Raffl, U. Stilla Photogrammetry and Remote Sensing, Technical University of Munich, 80290 Munich, Germany - [email protected], [email protected], [email protected], lukas.raffl@tum.de, [email protected] Commission II KEY WORDS: Bundle block adjustment, optical images, photogrammetric point cloud, change detection, crevices ABSTRACT: This contribution discusses the accuracy and the applicability of Photogrammetric point clouds based on dense image matching for the monitoring of gravitational mass movements caused by crevices. Four terrestrial image sequences for three different time epochs have been recorded and oriented using ground control point in a local reference frame. For the first epoch, two sequences are recorded, one in the morning and one in the afternoon to evaluate the noise level within the point clouds for a static geometry and changing light conditions. The second epoch is recorded a few months after the first epoch where also no significant change has occurred in between. The third epoch is recorded after one year with changes detected. As all point clouds are given in the same local coordinate frame and thus are co-registered via the ground control points, change detection is based on calculating the Multiscale-Model-to-Model-Cloud distances (M3C2) of the point clouds. Results show no movements for the first year, but identify significant movements comparing the third epoch taken in the second year. Besides the noise level estimation, the quality checks include the accuracy of the camera orientations based on ground control points, the covariances of the bundle adjustment, and a comparison the Geodetic measurements of additional control points and a laser scanning point cloud of a part of the crevice. Additionally, geological measurements of the movements have been performed using extensometers. 1. INTRODUCTION One of the direct consequences of climate change is the melt- ing of permafrost in alpine areas. In combination with extreme weather especially heavy rain falls gravitational mass movements in the mountains like rockfalls and hang slides are going to be a more frequently and more dangerous hazard (Clague and Stead, 2012). An important task in predicting these mass movements is the classification of the movement type based on a set of indicat- ors. Hungr et al. (2014) classifies five different movement types: fall, topple, slide, spread, and flow. This classification needs a detailed observation of movements. Here, observing the move- ment rate and direction of single points over time is not sufficient to model complex changes. This modeling requires the observa- tion of a various number of 3D points for longer time periods and leads to a classification of the movement type and an estimation of possible moving masses. Different measurement techniques are used for this purpose. Geological systems like extensometers (Malet et al., 2002; Mas- sey et al., 2013) are used to measure changes in the width of crevices. Geodetic measurements are mainly based on tachy- meter, GNSS receivers (Squarzoni et al., 2005), radar interfer- ometry or laser scanning (Roberti et al., 2018; Delacourt et al., 2007; Kasperski et al., 2010; Oppikofer et al., 2009). All these devices have in common their high costs compared to cameras and the need of professional personnel. Surveillance of larger areas or all potential risk areas in the Alps is more or less im- possible in that way. Using cameras allows handing over the job of data acquisition to non-professionals (Urban et al., 2019; Romeo et al., 2019; Mayr et al., 2019; Peppa et al., 2019; Kromer et al., 2019; Esposito et al., 2017; Hendrickx et al., 2019; Warrick * Corresponding author et al., 2016; Piermattei et al., 2016; Eltner et al., 2016; Stumpf et al., 2015; K¨ ab et al., 2014; Westoby et al., 2012). This contribution focuses on the assessment of possible ac- curacies in 3D reconstruction and change detection based on pho- togrammetric image sets. Photogrammetric results are compared to Geodetic measurements using tachymeter and laser scanner. 2. MEASUREMENT AND PROCESSING STRATEGY The evaluation of the accuracy and reliability in 3D reconstruc- tion and change detection and deformation analysis is split in two parts. In the first part, the accuracy of the 3D point generation is evaluated. The second part focuses on the change detection between two point clouds. A set of ground control points defines a local coordinate system. Camera orientations are unknown beforehand and the image set is oriented based on the ground control points. Two different sets of ground control points are used. Set one is fixed longterm ground control points defining the local coordinate system for registering the different epochs. Set two is ground control points that were individually measured for every epoch. The interior orientation of the camera is given and assumed to be fix for one epoch. 2.1 3D reconstruction The 3D reconstruction is based on the well-known two step process of bundle adjustment of the image based on tie point and ground control points (F¨ orstner and Wrobel, 2016) fol- lowed by a semi-global matching for dense point cloud gener- ation (Hirschmueller, 2008). Measurements of crevices are taken with an image overlap of at least 80%. The camera is moved ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume V-2-2020, 2020 XXIV ISPRS Congress (2020 edition) This contribution has been peer-reviewed. The double-blind peer-review was conducted on the basis of the full paper. https://doi.org/10.5194/isprs-annals-V-2-2020-687-2020 | © Authors 2020. CC BY 4.0 License. 687

Transcript of CHANGE DETECTION IN PHOTOGRAMMETRIC POINT ... - pf.bgu.tum… · change detection in...

Page 1: CHANGE DETECTION IN PHOTOGRAMMETRIC POINT ... - pf.bgu.tum… · change detection in photogrammetric point clouds for monitoring of alpine, gravitational mass movements a. dinkel,

CHANGE DETECTION IN PHOTOGRAMMETRIC POINT CLOUDS FORMONITORING OF ALPINE, GRAVITATIONAL MASS MOVEMENTS

A. Dinkel, L. Hoegner∗, A. Emmert, L. Raffl, U. Stilla

Photogrammetry and Remote Sensing, Technical University of Munich, 80290 Munich, Germany [email protected], [email protected], [email protected], [email protected], [email protected]

Commission II

KEY WORDS: Bundle block adjustment, optical images, photogrammetric point cloud, change detection, crevices

ABSTRACT:

This contribution discusses the accuracy and the applicability of Photogrammetric point clouds based on dense image matching for themonitoring of gravitational mass movements caused by crevices. Four terrestrial image sequences for three different time epochs havebeen recorded and oriented using ground control point in a local reference frame. For the first epoch, two sequences are recorded, onein the morning and one in the afternoon to evaluate the noise level within the point clouds for a static geometry and changing lightconditions. The second epoch is recorded a few months after the first epoch where also no significant change has occurred in between.The third epoch is recorded after one year with changes detected. As all point clouds are given in the same local coordinate frame andthus are co-registered via the ground control points, change detection is based on calculating the Multiscale-Model-to-Model-Clouddistances (M3C2) of the point clouds. Results show no movements for the first year, but identify significant movements comparingthe third epoch taken in the second year. Besides the noise level estimation, the quality checks include the accuracy of the cameraorientations based on ground control points, the covariances of the bundle adjustment, and a comparison the Geodetic measurementsof additional control points and a laser scanning point cloud of a part of the crevice. Additionally, geological measurements of themovements have been performed using extensometers.

1. INTRODUCTION

One of the direct consequences of climate change is the melt-ing of permafrost in alpine areas. In combination with extremeweather especially heavy rain falls gravitational mass movementsin the mountains like rockfalls and hang slides are going to be amore frequently and more dangerous hazard (Clague and Stead,2012). An important task in predicting these mass movements isthe classification of the movement type based on a set of indicat-ors. Hungr et al. (2014) classifies five different movement types:fall, topple, slide, spread, and flow. This classification needs adetailed observation of movements. Here, observing the move-ment rate and direction of single points over time is not sufficientto model complex changes. This modeling requires the observa-tion of a various number of 3D points for longer time periods andleads to a classification of the movement type and an estimationof possible moving masses.

Different measurement techniques are used for this purpose.Geological systems like extensometers (Malet et al., 2002; Mas-sey et al., 2013) are used to measure changes in the width ofcrevices. Geodetic measurements are mainly based on tachy-meter, GNSS receivers (Squarzoni et al., 2005), radar interfer-ometry or laser scanning (Roberti et al., 2018; Delacourt et al.,2007; Kasperski et al., 2010; Oppikofer et al., 2009). All thesedevices have in common their high costs compared to camerasand the need of professional personnel. Surveillance of largerareas or all potential risk areas in the Alps is more or less im-possible in that way. Using cameras allows handing over thejob of data acquisition to non-professionals (Urban et al., 2019;Romeo et al., 2019; Mayr et al., 2019; Peppa et al., 2019; Kromeret al., 2019; Esposito et al., 2017; Hendrickx et al., 2019; Warrick∗ Corresponding author

et al., 2016; Piermattei et al., 2016; Eltner et al., 2016; Stumpf etal., 2015; Kaab et al., 2014; Westoby et al., 2012).

This contribution focuses on the assessment of possible ac-curacies in 3D reconstruction and change detection based on pho-togrammetric image sets. Photogrammetric results are comparedto Geodetic measurements using tachymeter and laser scanner.

2. MEASUREMENT AND PROCESSING STRATEGY

The evaluation of the accuracy and reliability in 3D reconstruc-tion and change detection and deformation analysis is split in twoparts. In the first part, the accuracy of the 3D point generationis evaluated. The second part focuses on the change detectionbetween two point clouds.

A set of ground control points defines a local coordinate system.Camera orientations are unknown beforehand and the image set isoriented based on the ground control points. Two different sets ofground control points are used. Set one is fixed longterm groundcontrol points defining the local coordinate system for registeringthe different epochs. Set two is ground control points that wereindividually measured for every epoch. The interior orientationof the camera is given and assumed to be fix for one epoch.

2.1 3D reconstruction

The 3D reconstruction is based on the well-known two stepprocess of bundle adjustment of the image based on tie pointand ground control points (Forstner and Wrobel, 2016) fol-lowed by a semi-global matching for dense point cloud gener-ation (Hirschmueller, 2008). Measurements of crevices are takenwith an image overlap of at least 80%. The camera is moved

ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume V-2-2020, 2020 XXIV ISPRS Congress (2020 edition)

This contribution has been peer-reviewed. The double-blind peer-review was conducted on the basis of the full paper. https://doi.org/10.5194/isprs-annals-V-2-2020-687-2020 | © Authors 2020. CC BY 4.0 License.

687

Page 2: CHANGE DETECTION IN PHOTOGRAMMETRIC POINT ... - pf.bgu.tum… · change detection in photogrammetric point clouds for monitoring of alpine, gravitational mass movements a. dinkel,

Figure 1. Double-edge configuration.

Figure 2. Crevice configuration.

long the crevice in parallel view. The recording positions havebeen simulated beforehand with the calibrated camera. For largecrevices with several meters width, two different camera orienta-tion schemes are used:

• Double-edge configuration (Fig. 1): The images are orderedparallel to the crevice so that both edges of the crevice arevisible in every image. The position of the camera movesparallel to the pathway of the crevice.

• Crevice configuration (Fig. 2): The camera is tilted overthe edge of the crevice to record the scene in the interior ofthe crevice. Images now show only few or even no groundcontrol points and are connected to the first image set viatie points. Here, the camera is also moving parallel to thepathway of the crevice.

In such cases with large crevices, it is likely that the moving sideof the crevice is already unstable and cannot be accessed directly.For small crevices, both sides of the crevice are often accessible,but the crevice is to small to put the camera inside.

The bundle adjustment of the camera orientations uses imagepoints and ground control points as observations, the interior ori-entation is assumed to be fixed. The exterior orientation paramet-ers are the unknowns. As the collinearity equations (Forstner andWrobel, 2016) lead to a non-linear equation system in the bundleadjustment, an iterative adjustment has to be done and initial val-ues for the unknown parameters of the exterior orientation areestimated from the ground control points and their image points.

Three strategies are used to estimate the initial exterior orienta-tion. If at least six ground control points are visible, the direct lin-ear transform (DLT) is used (Luhmann, 2018). For three to fiveground control points, the minimal-object-information strategy(MOI) (Luhmann, 2018) is used. For images with less than threeground control points, the two neighboring images with estimatedexterior orientation are used to interpolate their exterior orienta-tion parameters linear from these two neighbors with the estim-ated image overlaps as weights. For five and four ground controlpoints in the MOI strategy, the optimal set of three ground con-trol points is found by searching the combination that spans thelargest triangle in the image.

The bundle adjustment uses variance components to model thedifferent accuracy levels of measured image points and measuredground control points. Random Sample Consensus (RANSAC)(Fischler and Bolles, 1981) removes outliers in the tie pointsearch before the bundle adjustment is performed. The resultingexterior orientations are the input for the semi-global matching(Hirschmueller, 2008).

2.2 Change detection and deformation analysis

Change detection consists of three major steps: i) coregistration;ii) noise detection; iii) detection of significant changes. As allimage sets are oriented based on the same set of ground controlpoints, one can assume the resulting point clouds to being alreadyregistered. An additional coregistration by distance minimizationlike Iterative Closest Point (Besl and McKay, 1992) would falsifythe change detection as it minimizes the distances of all pointsand thus would also include moved points in the registration.

The bundle adjustment calculates the inner quality of the resultingadjusted equation system. 3D points are only generated for thetie points and the ground control points. An absolute accuracy ofthe dense 3D point cloud is not calculated neither in the bundleadjustment nor in the semi-global matching. Estimating the noiselevel is very important to distinguish noise and real changes.

The proposed strategy uses the Multiscale-Model-to-Model-Cloud method (M3C2) (Lague et al., 2013) to find for every pointin one point cloud the corresponding neighbor in the other pointcloud and the distance between this point pair. M3C2 is used bothfor noise level estimation and change detection.

Given the exterior orientation of the images from the ground con-trol points, the registration accuracy is limited to the measure-ment accuracy of the ground control points. To refine the regis-tration we assume the area with the ground control points definingthe local coordinate system to be stable without changes. If so,a coregistration using Iterative Closest Point (ICP) (Fischler andBolles, 1981) leads to a refined registration of the point clouds.However, small geometric changes caused by the works on thesummit like changed position of equipment or small stone fallingdown during walking reduce the coregistration accuracy. Thesepoints are removed by an iterative process that generates a his-togram of M3C2 distances of corresponding points in the twopoint clouds. The five percent points with the highest distancesare cut out from the data set as points that either are wrong pointpairs or regions with unexpected small geometric changes. ICPis repeated for the new reduced data set and the M3C2 distancesare recalculated. This is repeated until all distances of the stablepart of the summit are below a threshold. The remaining pointcloud gives the parameters for the coregistration of the secondpoint cloud to the first. These transformation parameters are nowapplied on the original second point cloud including all points.

To estimate the noise level in this work, the next step comparesthe two point clouds generated from two image sets of the sameday taken with several hours difference. These point clouds in-clude changes in lighting conditions and slight differences in theorientations of the images. As no geometric changes around andin the crevice are expected for such a short period, both pointclouds should show M3C2 point distances in the level of the noiseof the point cloud generation quality. The median M3C2 distanceis the threshold of the noise level. The change detection uses theM3C2 distance with this threshold to remove noise and detect thegeometric changes.

ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume V-2-2020, 2020 XXIV ISPRS Congress (2020 edition)

This contribution has been peer-reviewed. The double-blind peer-review was conducted on the basis of the full paper. https://doi.org/10.5194/isprs-annals-V-2-2020-687-2020 | © Authors 2020. CC BY 4.0 License.

688

Page 3: CHANGE DETECTION IN PHOTOGRAMMETRIC POINT ... - pf.bgu.tum… · change detection in photogrammetric point clouds for monitoring of alpine, gravitational mass movements a. dinkel,

Figure 3. Summit of the Hochvogel mountain with the largecrevice.

Figure 4. Ortho image of the Hochvogel summit with thecrevices and geologic instruments marked.

3. EXPERIMENTS: THE HOCHVOGEL CREVICE

The mountain Hochvogel (Fig. 3) is situated at the borderbetween Germany (Bavaria) and Austria. Hochvogel is thehighest mountain in the Allgaeu alps with 2592 m. Hochvogelis characterized by its steep rock slopes and fast erosion pro-cesses caused by a very choosy limestone (Scholz and Scholz,1981; Krautblatter and Funk, 2010). The reduction of permafrostin the last decades destabilized the summit region of Hochvogel.This geologic behavior causes big mass movements and severalrock falls. The summit is crossed by a big crevice of up to sevenmeters width and visible down to 10 meters of depth. The deeperparts of the crevice are invisible due to the debris. It has a totallength of 35 meters through the whole summit from north east tosouth west and several narrow crevices in the south west of thesummit ( Fig. 4). The outer part of the summit that is separatingby the large crevice is moving in a gravitational movement to thesouth (Leinauer et al., 2019) with approx 2 cm per year. The 3 di-mensional complex movements in the summit of the Hochvogelshould be estimated.

In addition to the given ground control points used for definingthe local coordinate system and for the geodetic measurements(Fig. 5), additional ground control points have been placed in thescene to have more points for the initial orientation estimation.

Images have been taken during three measurement campaign on2018-07-09 (one image set in the morning, one image set in theafternoon for the noise level estimation), 2018-09-27, and 2019-07-19. A Sony Alpha 7RII with 42 megapixel and 15 mm focallength was used to take the images. The camera was mountedon a stick stand with a maximum length of 7 m (Fig. 6). Thecamera records the scene with at least 80% overlap in parallelviews moving along the crevice.

Figure 5. Ortho image of the Hochvogel summit with the fixedpermanent ground control points marked. Points ”G” are ground

control points for defining the local coordinate system. Points”K” are yellow balls as ground control points on the moving

rock side. Points ”R” are additional reflector marks.

Figure 6. Camera mounted on the stick stand recording thecrevice in the right part of the image.

Table 1 lists the number of images, visible ground control pointsand manually added TIE points for the three measurement cam-paigns.

Campaign images GCPs manual TIEJuly 2018-A 183 32 20July 2018-B 106 32 20

July 2018 407 32 20Sept 2018 645 55 47July 2019 510 74 55

Table 1. Different measurements: July 2018-A and B: Same daymeasurements for noise level estimation. July 2018, Sept 2018,

July 2019: Change detection measurements.

4. RESULTS AND DISCUSSION

During the measurements, several groups from geology, geodesyand photogrammetry were working on the summit and installingground control points and equipment. This leads to small geo-metric changes and changing occlusions by person during the re-cordings.

ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume V-2-2020, 2020 XXIV ISPRS Congress (2020 edition)

This contribution has been peer-reviewed. The double-blind peer-review was conducted on the basis of the full paper. https://doi.org/10.5194/isprs-annals-V-2-2020-687-2020 | © Authors 2020. CC BY 4.0 License.

689

Page 4: CHANGE DETECTION IN PHOTOGRAMMETRIC POINT ... - pf.bgu.tum… · change detection in photogrammetric point clouds for monitoring of alpine, gravitational mass movements a. dinkel,

Geo XY/Z X Y Z errorR1 0.004/ 0.002 -0.003 0.002 0.001 0.432R2 0.004/ 0.002 -0.001 0.001 0.001 0.364R3 0.004/ 0.002 -0.000 -0.000 0.000 0.425R5 0.004/ 0.002 0.001 0.001 0.000 0.502R4 0.004/ 0.002 -0.001 -0.001 -0.000 0.431R6 0.004/ 0.002 0.001 -0.002 0.001 0.546R7 0.004/ 0.002 -0.001 -0.002 -0.000 0.526

Table 2. Standard deviation of July 2018 measures of GroundControl points Geodesy (Geo XY/Z in [m]) compared to

Photogrammetry (X,Y,Z in [m]) and reprojection error of groundcontrol points in [pixel].

Figure 7. July 2018 data set: Accuracy of the 3D objectcoordinates after the bundle adjustment.

4.1 3D reconstruction

Table 2 shows the accuracy of the ground control points definingthe local coordinate system measured from geodetic instruments(Wunderlich et al., 2019) and the estimated standard deviationsof the ground control points after the bundle adjustment togetherwith the remaining mean reprojection error of the ground controlpoints in the images. One can see that the accuracies are in arange of a few millimeters.

The results for the 3D object coordinates estimated in the bundleadjustment shows that the standard deviation of the points is be-low 1 cm in the center of the image set and gets worse to theborders of the set (Fig. 7 and 8. This is mainly caused by thelower number of images that see these points and by the config-uration of the ground control points. The ground control pointsare aligned along an axis due to the formation of the crevice andaccessibility of the different parts of the summit. This makes theorientation unstable to a certain point. The reprojection error forthe three image sets for July 2018, September 2018, and July2019 are 0, 225pixel, 0, 119pixel and 0, 162pixel.

Looking at figure 9, one can see the M3C2 distances of the photo-grammetric point cloud from September 2018 to the correspond-ing point cloud taken with a terrestrial laser scanner. The standarddeviation of the scanner σ is about 2mm. The mean M3C2 dis-tance of the point clouds is 0.94mm with a standard deviation of5.6mm. As seen in the figure 9, the main difference is in a smallarea where different occlusions of scanner and camera may cause

Figure 8. July 2018 data set: Accuracy of the point coordinatesin the XY plane after the bundle adjustment.

Figure 9. July 2018 data set: Distance of terrestrial laser scannerpoint cloud to photogrammetric point cloud. Both point clouds

are registered using the ground control points.

changes. Similar reasons are the small high differences which aremainly in the shadow areas of rocks.

4.2 Noise level estimation

The noise level estimation uses the two image sets July 2018 Aand B taken during the first measurement campaign and com-pares the two generated point clouds. As we assume that thereis no change in the geometry of the summit, the point clouds arecoregistered using ICP to remove possible errors from the limitedaccuracy of the ground control point measurements. M3C2 cal-culates the distances of point pairs of the two point clouds. Themean distance µ of the points is 5.35 ∗ 10−4m. The standarddeviation σ is 0.013m.

4.3 Change detection

In change detection, two comparisons are done. The first com-pares the point cloud from epoch one (July 2018) to the second(September 2018) where no movements have been detectedneither by geological nor geodetic measurements. In fact, us-ing the estimated noise level, almost no significant changes aredetected in the photogrammetric point clouds (Fig. 10). The leftimages shows the detected changes. Small rocks have moved dur-ing the summer, but no movement of the whole crevice is visible.The right images shows the cleaned point cloud with stable pointsonly that have been used for the fine coregistration.

The mean value µ of the M3C2 distances is 9.5 ∗ 10−5m andthe standard deviation σ is 0.0015m. This is below the estimatednoise level and thus no movement is detected.

ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume V-2-2020, 2020 XXIV ISPRS Congress (2020 edition)

This contribution has been peer-reviewed. The double-blind peer-review was conducted on the basis of the full paper. https://doi.org/10.5194/isprs-annals-V-2-2020-687-2020 | © Authors 2020. CC BY 4.0 License.

690

Page 5: CHANGE DETECTION IN PHOTOGRAMMETRIC POINT ... - pf.bgu.tum… · change detection in photogrammetric point clouds for monitoring of alpine, gravitational mass movements a. dinkel,

Figure 10. Changes between July and September 2018. Left:detected changes, Right: Cleaned point cloud with remaining

stable points for coregistration

Figure 11. Changes between September 2018 and July 2019.M3C2 distances of up to 3 cm (red) have been detected.

Figure 12. Changes between September 2018 and July 2019.Histogram of the M3C2 distances. Three peaks can be seen in

the distance histogram.

If we have a look at the distances of the point clouds fromSeptember 2018 to July 2019 (Fig. 11, the results look totallydifferent. On can see M3C2 differences of up to 3 cm in the leftpart of the crevice and that more or less all parts of the creviceare moving. If one looks at the histogram of the M3C2 distances(Fig. 12, one can see three peaks at -1.4 cm, +8 mm, and +1.6cm. If the points are segmented using a histogram based thresholdaround these three peaks, one can see, that all three peaks belongto a certain characteristic (Fig. 13. The 8mm peak is more orless only on horizontal plains that are going down, whereas the1.6 cm peak is located on more or less vertical surfaces that aremoving away from the stable part of the summit. This distancesand their locations describe movements of the sliding part of thesummit. The peak at -1.4 cm belongs to objects that are fallinginto to crevice. The movement in total seems to be stronger onthe left side (Fig. 9. Characterising the type of movement, theresults indicate a combination of topple and slide which fits tothe geologic reference measurements.

An interesting observation can be made at the smaller crevicethat is perpendicular to the big crevice. Here, we see a movementalong the crevice, more or less in the same direction as the move-ment in the main crevice. The distances in the point clouds ofSeptember 2018 to July 2019 (Fig. 15 at first show an unstruc-tured change. But, if one adds virtual connection lines of objectsin the September 2018 data set and compares this to the July 2019

Figure 13. Changes between September 2018 and July 2019.M3C2 distances grouped using histogram based thresholds.

Peak - 1.4 cm describes movements of objects towards the stablepart. Peak 8 mm describes horizontal plains that are moving

down. Peak 1.6 cm fits to vertical surfaces moving away fromthe stable part of the summit.

Figure 14. Changes between September 2018 and July 2019.Differences in the small side crevice.Blue arrow shows

movement direction. Lines show significant structures that havemoved.

data set one can see a displacement of these characteristic struc-tures.

If we compare our measurements and the interpretation with thegeodetic measurements (Fig. 14, we can see that the changes ofthe control points on the moving part of the summit support the

ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume V-2-2020, 2020 XXIV ISPRS Congress (2020 edition)

This contribution has been peer-reviewed. The double-blind peer-review was conducted on the basis of the full paper. https://doi.org/10.5194/isprs-annals-V-2-2020-687-2020 | © Authors 2020. CC BY 4.0 License.

691

Page 6: CHANGE DETECTION IN PHOTOGRAMMETRIC POINT ... - pf.bgu.tum… · change detection in photogrammetric point clouds for monitoring of alpine, gravitational mass movements a. dinkel,

Figure 15. Changes between September 2018 and July 2019.Geodetic measurement of movement of control points.

results of the photogrammetric measurement and interpretationof the point cloud differences. The mean value of the movementis 2cm which fits to our estimation of the movement of the ver-tical surfaces away from the stable part of the summit. Table 3shows the comparison of the geodetic measured movement andthe estimated movement from the point cloud differences.

5. CONCLUSION AND OUTLOOK

The image acquisitions in this project were done with a terrestrialcamera system. It is possible to replace this terrestrial camera bya UAV/RPAS mounted camera system. This would most likelyspeed up the recording process that was quite complex due to sev-eral groups working at the summit at the same time and thus gen-erating artificial changes from persons and material that had to beremoved before processing. A second limitation is the possibleground control points configuration. As the summit and crevicehave an elongated shape and the moving part is hardly access-ible, the ground control points for the coordinate system are alsoelongated which makes the registration unstable at the ends of theimage sets. Nevertheless, it is possible to reach high geometricaccuracies for the photogrammetric point clouds. If one considersthe real noise level instead of the inner accuracy of the bundle ad-justment and dense matching, it is clear that relevant movementscan be detected. To reach this accuracy, it is possible to have

GCP dx dy dz 2D pose distance point cloudR1 1,3 1,7 0,2 2,14 2,15 2,20R2 1,3 1,7 0,1 2,14 2,14 1,99R3 0,9 1,6 0,6 1,84 1,84 1,78R5 1,1 1,8 0,1 2,11 2,13 1,47R6 1,2 1,8 0,1 2,16 2,17 2,35R7 0,9 1,7 0,1 1,92 1,93 1,66K1 1,3 2,0 0,5 2,39 2,44 2,11K2 0,8 1,7 0,0 1,88 1,88 2,18K3 1,2 1,9 0,0 2,25 2,25 2,08K4 0,7 1,8 0,2 1,93 1,94 1,87

Table 3. Comparison of geodetic measured movement andestimated movement from point cloud differences in [cm].

a good ground control point set for absolute image orientation.Compared to the terrestrial laser scanner used in parallel we cansee that we reach similar accuracies. The overall recording timesare also comparable as the laser scanner needs a couple of pos-itions to be placed for observing the whole crevice. Regardingthe view into the crevice itself, Photogrammetry has the bene-fit of being able to looking down to areas occluded for the laserscanner. Airborne laser scanning systems however could reducethat position limitations of the terrestrial scanner. Mounted onaircraft, the point density will be much lower than for the photo-grammetric point cloud, for UAV mounted systems the accuraciesand point densities are limited. So, we propose a combinationof low attitude photogrammetric UAV-based recordings and, foroccluded areas, terrestrial images. Image sequences would alsoallow for image based change detection. This could be taken intoaccount for additional change detection, but is limited due to itsdependency on lighting conditions, especially shadow corners.

Future steps include the use of UAV/RPAS systems for carryingthe camera. This has the advantage that exterior orientation para-meters are measured for every image and can be used as initialvalues in the bundle adjustment. Nevertheless, using only theseorientation will only lead to an optimized inner accuracy. It hasto be investigated, whether this accuracy in combination with thenoise removal and an point based coregistration can reach thesame absolute accuracy as the method based in ground controlpoints.

The interpretation of changes as specific movements can be im-proved by adding a segmentation to the point cloud and grouppoints by radiometric and geometric features and detected edgesas well as similar estimated movement vectors. This way, itwould be possible, to track objects instead of assigning pointsof two point clouds to each other that show not necessarily thesame object.

References

Besl, P. J. and McKay, N. D., 1992. Method for registration of3-d shapes. In: Sensor fusion IV: control paradigms and datastructures, Vol. SPIE 1611, International Society for Optics andPhotonics, pp. 586–606.

Clague, J. J. and Stead, D., 2012. Landslides: types, mechanismsand modeling. Cambridge University Press.

Delacourt, C., Allemand, P., Berthier, E., Raucoules, D., Casson,B., Grandjean, P., Pambrun, C. and Varel, E., 2007. Remote-sensing techniques for analysing landslide kinematics: a review.Bulletin de la Societe Geologique de France 178(2), pp. 89–100.

Eltner, A., Kaiser, A., Castillo, C., Rock, G., Neugirg, F.and Abellan, A., 2016. Image-based surface reconstruction ingeomorphometry–merits, limits and developments. Earth Sur-face Dynamics 4(2), pp. 359–389.

Esposito, G., Salvini, R., Matano, F., Sacchi, M., Danzi, M.,Somma, R. and Troise, C., 2017. Multitemporal monitoring ofa coastal landslide through sfm-derived point cloud comparison.The Photogrammetric Record 32(160), pp. 459–479.

Fischler, M. and Bolles, R., 1981. Random sample consensus: Aparadigm for model fit-ting with applications to image analysisand automated cartography. Communications of the ACM 24(6),pp. 381–395.

Forstner, W. and Wrobel, B. P., 2016. Photogrammetric Com-puter Vision. Statistics, geometry, Orientation and Reconstruc-tion. Springer, Berlin.

ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume V-2-2020, 2020 XXIV ISPRS Congress (2020 edition)

This contribution has been peer-reviewed. The double-blind peer-review was conducted on the basis of the full paper. https://doi.org/10.5194/isprs-annals-V-2-2020-687-2020 | © Authors 2020. CC BY 4.0 License.

692

Page 7: CHANGE DETECTION IN PHOTOGRAMMETRIC POINT ... - pf.bgu.tum… · change detection in photogrammetric point clouds for monitoring of alpine, gravitational mass movements a. dinkel,

Hendrickx, H., Vivero, S., De Cock, L., De Wit, B., De Maeyer,P., Lambiel, C., Delaloye, R., Nyssen, J. and Frankl, A., 2019.The reproducibility of sfm algorithms to produce detailed digitalsurface models: the example of photoscan applied to a high-alpine rock glacier. Remote sensing letters 10(1), pp. 11–20.

Hirschmueller, H., 2008. Stereo processing by semiglobal match-ing and mutual information. IEEE Transactions on Pattern Ana-lysis and Machine Intelligence 30(2), pp. 328–341.

Hungr, O., Leroueil, S. and Picarelli, L., 2014. The varnes classi-fication of landslide types, an update. Landslides 11(2), pp. 167–194.

Kaab, A., Girod, L. M. R. and Berthling, I. T., 2014. Surface kin-ematics of periglacial sorted circles using structure-from-motiontechnology. The Cryosphere 8, pp. 1041–1056.

Kasperski, J., Delacourt, C., Allemand, P., Potherat, P., Jaud, M.and Varrel, E., 2010. Application of a terrestrial laser scanner (tls)to the study of the sechilienne landslide (isere, france). RemoteSensing 2(12), pp. 2785–2802.

Krautblatter, M. and Funk, D., 2010. A rock-/ice mechanicalmodel for the destabilisation of permafrost rocks. In: EGU Gen-eral Assembly Conference Abstracts, Vol. 12, p. 4071.

Kromer, R., Walton, G., Gray, B., Lato, M. et al., 2019. Develop-ment and optimization of an automated fixed-location time lapsephotogrammetric rock slope monitoring system. Remote Sensing11(16), pp. 1890.

Lague, D., Brodu, N. and Leroux, J., 2013. Accurate 3d compar-ison of complex topography with terrestrial laser scanner: applic-ation to the rangitikei canyon (n-z).

Leinauer, J., Jacobs, B. and Friedrich, B., 2019. Process dynam-ics and early warning strategies for a preparing highmagnituderock slope failure at the hochvogel (allgau alps). In: EGU Gen-eral Assembly Conference Abstracts, Vol. 21, p. 2196.

Luhmann, T., 2018. Nahbereichsphotogrammetrie – Grundlagen,Methoden und Anwendungen. 4, vollig neu bearbeitete und er-weiterte Auflage. Wichmann VDE Verlag, Berlin und Offenbach.

Malet, J.-P., Maquaire, O. and Calais, E., 2002. The use of globalpositioning system techniques for the continuous monitoring oflandslides: application to the super-sauze earthflow (alpes-de-haute-provence, france). Geomorphology 43(1-2), pp. 33–54.

Massey, C. I., Petley, D. N. and McSaveney, M., 2013. Patternsof movement in reactivated landslides. Engineering Geology 159,pp. 1–19.

Mayr, A., Bremer, M., Rutzinger, M. and Geitner, C., 2019.Unmanned aerial vehicle laser scanning for erosion monitoringin alpine grassland. ISPRS Annals of Photogrammetry, RemoteSensing and Spatial Information Sciences pp. 405–412.

Oppikofer, T., Jaboyedoff, M., Blikra, L., Derron, M.-H. andMetzger, R., 2009. Characterization and monitoring of the aknesrockslide using terrestrial laser scanning. Natural Hazards andEarth System Sciences 9(3), pp. 1003–1019.

Peppa, M. V., Mills, J. P., Moore, P., Miller, P. E. and Chambers,J. E., 2019. Automated co-registration and calibration in sfmphotogrammetry for landslide change detection. Earth SurfaceProcesses and Landforms 44(1), pp. 287–303.

Piermattei, L., Carturan, L., de Blasi, F., Tarolli, P.,Dalla Fontana, G., Vettore, A. and Pfeifer, N., 2016. Suitabilityof ground-based sfm–mvs for monitoring glacial and periglacialprocesses. Earth Surface Dynamics 4(2), pp. 425–443.

Roberti, G., Ward, B., de Vries, B. v. W., Friele, P., Perotti,L., Clague, J. J. and Giardino, M., 2018. Precursory slope dis-tress prior to the 2010 mount meager landslide, british columbia.Landslides 15(4), pp. 637–647.

Romeo, S., Di Matteo, L., Kieffer, D. S., Tosi, G., Stop-pini, A. and Radicioni, F., 2019. The use of gigapixelphotogrammetry for the understanding of landslide pro-cesses in alpine terrain. Geosciences 9(2), pp. 1–99, ht-tps://doi.org/10.3390/geosciences9020099.

Scholz, H. and Scholz, U., 1981. Das Werden der Allgauer Land-schaft. Eine kleine Erdgeschichte des Allgaus. Verlag fur Heimat-pflege, Kempten.

Squarzoni, C., Delacourt, C. and Allemand, P., 2005. Differen-tial single-frequency gps monitoring of the la valette landslide(french alps). Engineering Geology 79(3-4), pp. 215–229.

Stumpf, A., Malet, J.-P., Allemand, P., Pierrot-Deseilligny, M.and Skupinski, G., 2015. Ground-based multi-view photogram-metry for the monitoring of landslide deformation and erosion.Geomorphology 231, pp. 130–145.

Urban, R., Stroner, M., Blistan, P., Kovanic, L., Patera, M., Jacko,S., Duriska, I., Kelemen, M. and Szabo, S., 2019. The suitabilityof uas for mass movement monitoring caused by torrential rain-fall—a study on the talus cones in the alpine terrain in high tatras,slovakia. ISPRS International Journal of Geo-Information 8(8),pp. 1–18, https://doi.org/10.3390/ijgi8080317.

Warrick, J. A., Ritchie, A. C., Adelman, G., Adelman, K. andLimber, P. W., 2016. New techniques to measure cliff changefrom historical oblique aerial photographs and structure-from-motion photogrammetry. Journal of Coastal Research 33(1),pp. 39–55.

Westoby, M. J., Brasington, J., Glasser, N. F., Hambrey, M. J. andReynolds, J. M., 2012. ‘structure-from-motion’photogrammetry:A low-cost, effective tool for geoscience applications. Geomor-phology 179, pp. 300–314.

Wunderlich, T., Raffl, L. and Wiedemann, W., 2019.Wiedererkennung–zwei losungen fur die strenge deformationsan-alyse flachenhafter beobachtungen der ingenieurgeodasie. In:K. Hanke and T. Weinold (eds), 20. Internationale GeodatischeWoche Obergurgl 2019,“Arbeitskreis Rutschung, Setzung, De-formation”, pp. 264–273.

ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume V-2-2020, 2020 XXIV ISPRS Congress (2020 edition)

This contribution has been peer-reviewed. The double-blind peer-review was conducted on the basis of the full paper. https://doi.org/10.5194/isprs-annals-V-2-2020-687-2020 | © Authors 2020. CC BY 4.0 License.

693